top of page
Writer's pictureFinHacksQuants

FINHACKSQUANTS.COM @HK FinTech Week Pitch 20191103

Updated: Nov 7, 2019




B4 AND After NONE|FORESEEABLE
Problem Statement
Kelvin and Wayne @FinHacks Hackathon 2018
SEP 2019 FinHacksQuants.com approved by Google Startup

10thDEC2004 Kelvin Awarded [HKAYP Gold] by HKCE Mr. TUNG Chee-Wah

Founder Introduction: LinkedIn


Q&A : How many of you ladies and gentlemen do look at the PRICE b4 u buy or sell anything? please raise your hands up hold it up for couple seconds for some statistics hands counts photo taken first please.在座各位貴賓有邊位買賣嘢當時會睇價格的,請舉手幾秒統計影張相。


World's 1st fully automated 1440P Ultrawide HD LIVE #AI Observatory an emerging FinTech solution focusing in [Global AI Logic] tackling global problems such as:

1. diminishing investment returns,

2. meeting demands for tangible HD visuals on what is Forecastifiable in immediate future, 3. a way out for local institutions/brokerages to escape "Zero Commission Models" that will result in a race to the bottom, 

by our proprietary Global AI+Maths Price Forecasts Models(~Weather Forecasts)


HOW IT WORKS?

1. [GREEN=RISE👆* RED=FALL👇]

2. [YELLOW/LAVENDER =FLAT🙌]

3. [LEFT   Side of Vertical Black Line= HISTORICAL👶👧]

4. [RIGHT Side of Vertical Black Line= FORECASTS 👴👵]


Possible Answers: these (False|Positive or True|Negative) kinds of Breakups/Breakdowns could means just to fill up all those Limit Stop Orders hanging around support/resistant levels| these could be smart money/ dumb money | unknown fresh breaking news | fat finger | insiders positions | institutions' positions | Black pool | OTC transactions.......etc .

Engineers need to decipher those respective NLP token and time-stamps' vectorized outputs and trace back all those possible neurons paths to locate where raw data came from and what respective labels are, then log it. You are invited to further checkout two exclusive new terms [Think-to-Trade] and [Matrix-Decipher-er] in a 4 min long video. (Vice Versa above for the other way around ie. BreakDowns/Breakups)


**************************************************************


Business Intelligence: These NLP outputs are then sent to the backbones Global AI for analysis, final directions to be output on the RIGHT side of the Vertical Black Line, we name these right side outputs as "FORECASTIFY".NLP Outputs are useful in some portions of our algo's logic flows : For example, when underlying price sharply broke up penetrated through major resistant levels in a morning OPEN full of Negative Market News!


Reactions - Authorize Algos some Possible auto-reactions: In emergency consider Closing All Positions | Increase insurance positions | decrease underlying positions | no new entry period | Artificial fat finger(joking).....etc


AI Chat Bots : User Chat Bot Inquiry "Price sharply broke up, what the heck is happening?" Chat Bots would reply users inquiries by adjusting results based on user's existing behaviour and relevance. Then analyze user's feedback again for relevance and confidence.


What Human probably can do? Human|Algos|Liquidity providers|HFT Markets Maker should react to their positions, for example Hedge|Unwind|Tune up/down vega|Widen/narrower spreads|Pour in/Dry up liquidity, against positions in time when necessary. When in emergency dumping positions legally.




Base on the foundations as has been discussed by Cheney and Medhy at EY Wavespace on 20191028. Our approach is similar, to deploy sentiments insights via BERT as parts of the Price Forecasts to be considered inside the Forecastify (right side of the vertical black line) visuals basing on NLP from (for example) #News(Bloomberg|CNBC), #Social Media(Twitter), or even #Big Data(Google Analytics on our Website,comments from YouTube).


Technical|Logical Summary: To utilise any set of data, there are needs for Data preparations: Data Collection, Selenium Data Scrapings, Cleansing, EXCEL VBA templates, SQL, Structured|UnStructured|Semi-Structured, Positive|Negative keywords, tokenization, vectorization, neural network, encoder, hidden states, weightings, biases. labeled|unlabeled inputs, supervised|unsupervised ML to self-generate|destroy of tokens/hidden layers Bi-directionally, data lakes......


/*(Technical|Logic Can Skip if no time) ***************************Skipped**************************************(Technicals|Logic Can Skip if no time)*/


102 views0 comments

Comments


bottom of page